THE FINALITY PROJECT PRESENTATION
The two top challenges facing the telecom industry in the next five to ten years are the lack of a skilled work-force and ensuring network quality and reliability. These challenges are in fact tightly coupled due to the rapid change in the telecom landscape, caused by the pervasive adoption of softwarization and virtualization.
Softwarization and virtualization are key to future financial and environmental sustainability in the telecom sector, as they provide increased flexibility and lower life cycle management costs. At the same time, increased flexibility comes at the price of unprecedented complexity.
Developing and managing this new generation of softwarized telecom infrastructure requires a workforce skilled in software and, importantly, skilled in a new generation of resource allocation (RA) algorithms that can ensure network quality and reliability for the entire gamut of complex services in 6G systems.
A combination of twinning, i.e., maintaining dynamic models of the physical environment used for offline training prior to deployment, and the use of machine learning (ML) is considered the ideal foundation for building a new generation of AI-based RA algorithms.
These solutions have the potential to achieve the ambitious sustainability objectives of the Telecom industry through tailored data-driven optimization tools, and enable the efficient integration of renewable energy sources.
Nonetheless, for this approach to be feasible in a critical infrastructure, new ML algorithms have to be developed that are safe, i.e., to respect operational constraints, as the potential consequences of incorrect decisions could have, e.g., unprecedented socio-economic impact.
At the same time, safety requires a high degree of promptness to meet real-time constraints of production environments. Algorithms have to adapt promptly to changing environmental conditions and failures.
Our mission
The FINALITY doctoral program is set out to address this incumbent challenge by developing a new educational curriculum that combines ML with sectoral skills in the telecom domain, complemented by transferable skills
Each trained doctoral candidate (DC) will be equipped with the core knowledge needed to pave the way to a responsible transition towards the full autonomy of critical systems beyond current human supervised operations. The doctoral team will develop an AI toolkit for prompt and safe RA, an asset in the wish-list of most industrial sectors nowadays aiming at maintaining human-in-the-loop control. FINALITY will address a high-risk use case of AI in critical infrastructures: the Communications Sector, which is a cornerstone of a sovereign EU economy, underlying the operations of all businesses, public safety organizations and governments, providing an “enabling function” across all critical infrastructures. In this context, the FINALITY trainees will develop methods that ensure that constraints and rules are embedded, by design, in the learning phase of ML algorithms.
They will be exposed to application use-cases during secondments in leading EU companies of the sector, and make their results available across all areas of critical infrastructure systems, bringing significant benefits including, e.g., service guarantees under limited resources resulting in reduced injuries and fatalities due to accidents. The EU puts a strong focus on addressing the skills gap in AI.
The profile of FINALITY trainees on safe ML for RA appears on the critical path to attain the EU approach to AI. While aligning to the future EU regulatory framework for AI, the DCs will also foster the “reinforcement of the skills” agenda, an inclusive program aimed at the green and digital transformation of the EU economy. To this aim, the fairness of AI methods and the emphasis on parsimonious mathematical tools with reduced computational footprint will be a pivotal KPI of the solutions sought after by FINALITY DCs.